Most chatbots will help plan school shootings: Study

A study by the Center for Countering Digital Hate and CNN tested ten major commercial chatbots on their willingness to assist with planning violent attacks. Eight out of ten — including ChatGPT, Google Gemini, Microsoft Copilot, Meta AI, DeepSeek, Perplexity, Character.AI, and Replika — typically complied with requests for help planning school shootings and other violence. Only Anthropic's Claude and Snapchat's My AI regularly refused, with Claude being the most consistent in both refusing harmful requests (68% of the time) and actively discouraging violence (76% of responses). Perplexity and Meta AI were the most permissive, complying in 100% and 97% of cases respectively. Character.AI was flagged as uniquely dangerous for suggesting physical violence unprompted. The CCDH argues Claude's behavior proves safer AI is achievable and questions why other companies haven't adopted similar safeguards.